Exploring the link between spectra, inherent optical properties in the water column, and sea surface temperature and salinity

Abstract

Sea surface salinity and temperature are important measures of ocean health. They provide information about ocean warming, atmospheric interactions, and acidification, with further effects on the global thermohaline circulation and as a consequence the global water cycle. In coastal waters they provide information about sub mesoscale circulations and tidal currents, riverine discharge and upwelling effects. This paper explores the methodology to extract sea surface salinity (SSS) and temperature (SST) from ground based hyperspectral ocean radiance. Water leaving radiance is linked to the inherent optical properties of the water column, effected by the constituent parts. Hyperspectral data at ground level is then used as input to train a linear regression model against temporally and spatially matched water data of SSS and SST. Furthermore, a neural network model to be able to estimate the SST and SSS with the hyperspectral data averaged to multispectral bands to emulate the satellite use case. The neural network model is able to learn the relationship between the multispectral radiance to both SSS and SST values, and can predict these with a root mean square error (RMSE) of 0.2PSU and 0.1 degree respectively. This demonstrates the feasibility of similar algorithms applied to multispectral ocean colour satellites with enhanced coverage and spatial resolution.